Image analysis for malaria parasite detection from microscopic images of thick blood smear

Malaria is a major health issue and causes millions of deaths a year worldwide. Since the diagnosis of malaria is predominately done using light microscopy method, well-trained microscopists are required. By using thick blood smear, a large amount of blood can be examined quickly and easily. This work deals with the automatic estimation of parasite density in `parasites per microliter of blood' from the microscopic images of Giemsa-stained thick blood smear. The algorithm is primarily divided into three steps: (1) preprocessing & segmentation, (2) feature extraction and (3) classification. In the preprocessing step, an attempt is made to reduce variations due to various factors like lighting conditions and concentration of staining solution. The image is segmented using adaptive thresholding, followed by several mathematical morphological operations. In the second step, various features based on shape, texture, color and frequency domain are extracted. Using the classification step, the parasite candidate is classified into its correct life stage or classified as leukocytes. The novelty of the algorithm is that it can detect all the life stages (ring, trophozoite, schizont, gametocyte) of parasites and leukocytes unlike detecting only ring life stage in the state-of-the-art algorithms. The discrepancy in the automated parasite count by the proposed algorithm is 7.14%, which is suitable for computer aided diagnosis (CAD) of malaria according to world health organization (WHO) quality control standards.